[R-gui] Re: R-SIG-GUI Digest, Vol 20, Issue 7

Philippe,

I think your impressions are quite accurate: experts on statistics are more 
interested on programming statistical problems than user interfaces, and 
non experts would like to have GUI's but they do not have the skills.
You may think that this problem could be solved technically. If  tools for 
programming graphical interfaces were available, experts might use them to 
produce good GUI's. However, technique only is probably not enough. It is 
necessary a culture that promotes this behaviour.
We might learn from past similar experiences. An statistical environment 
that enjoyed certain success before R came was Lisp-Stat. Lisp-Stat had 
superb user interface capabilities at the time if came up. Capabilities 
that, nowadays, are probably not very impressive but that, if available in 
R, would be sufficient for many purposes. However, experience showed that 
the people programming in Lisp-Stat very often did not put as much effort 
on the graphical interface as they put on the statistical programming. 
Lisp-Stat is now basically dead but we still can learn something from it: 
programmers may not program user interfaces even though they have the tools.
My impression is that commercial programs will continue leading the GUI 
efforts in the close future. Free software written by statisticians 
interested on statistics will not have the bells and whistles we are used 
to see in commercial software in the close future. After all, software 
companies estimate that the programming effort of  the GUI of a typical 
application is about the 60% or 80% of the program. Only paid people is 
probably to be very excited about carrying out this effort.
The only solution I see for the problem is a change in attitude about the 
importance of  the problem of user interfaces. This change of attitude 
should lead to increasing the reward available at academic places for 
working in this area. After all, at this moment, most of the people see the 
statistical part as the "ice cream" and the interface as the "topping" of 
statistical systems. In other words, not many people believe that the 
interfaces, GUI or not, suppose an academic challenge that deserve academic 
reward. As a consequence, there are not journals, books or conferences 
discussing methods, techniques, case stories or comparisons of features of 
people trying to develop good interfaces. The developments on this area do 
not accumulate. Hence, if the attitude about developing interfaces does not 
change, it is very difficult that many people will get involved on 
programming them, and the progress is going to be very limited.
You might remember the times were programming statistical software was a 
second class activity and it did not deserve academical reward. Now, things 
have changed and programmers of statistical software enjoy great reputation 
for carrying out their work. So, books and papers on statistical systems 
are among the most cited and read by everybody. Journals like JCGS, 
encouraging the publication of papers with such contents have contributed 
to such state of things. Who knows, programming of good interfaces might 
reach a similar status some day.